基于LightGBM的LAA/Wi-Fi混合网络的竞争窗口自适应预调整  

Adaptive Preliminary Adjustment of Contention Window of LAA/Wi-Fi Hybrid Network Based on LightGBM

作　　者：冯义磊 唐震洲 FENG Yilei;TANG Zhenzhou(College of Computer Science and Artificial Intelligence,Wenzhou University,Wenzhou,China 325035)

机构地区：[1]温州大学计算机与人工智能学院,浙江温州325035

出　　处：《温州大学学报（自然科学版）》2021年第1期49-58,共10页Journal of Wenzhou University(Natural Science Edition)

基　　金：浙江省自然科学基金重点项目(LZ20F010008);温州市基础性科研项目(G20180008)。

摘　　要：授权频谱辅助接入技术(License Assisted Access,LAA)可以把LTE网络部署在未授权频段上.为了与Wi-Fi网络友好共存,LAA网络采用先听后说(Listen Before Talk,LBT)机制.但是,LAA和Wi-Fi共存网络的性能很大程度上取决于LBT中的退避机制,而竞争窗口的增加则是由当前传输机会(Transmission Opportunity,TXOP)的混合自动重传请求(Hybrid Automatic Repeat Request,HARQ)反馈情况来决定的.为此,本文首先利用LightGBM等6种机器学习算法,对TXOP第一个子帧的HARQ反馈的NACK占比是否达到80%进行了预测,由于LightGBM算法91.6%的准确率和0.98的AUC值(Area Under Curve),故选其作为实时预测模型.结果表明,采用了本文所提出的竞争窗口自适应预调整机制的LAA/Wi-Fi混合网络相比于没有采用该机制的LAA/Wi-Fi混合网络,吞吐量提升了21.266%.License Assisted Access(LAA)enables the deployment of LTE network in unlicensed frequency band.In order to coexist friendly with Wi-Fi network,LAA network adopts Listen before Talk(LBT)mechanism.However,the performance of the network with LAA and Wi-Fi is largely determined by the backoff mechanism in LBT.For backoff mechanism,in addition,the increase of the contention window is determined by the feedback of hybrid automatic repeat request(HARQ)of the current Transmit Opportunity(TXOP).In view of this,this paper first uses six machine learning algorithms,such as lightgbm,to predict whether the proportion of NACK of HARQ feedback in the first subframe of TXOP reaches 80%.Meanwhile,LightGBM algorithm is selected as a real-time prediction model because of its 91.6%accuracy and its 0.98 AUC value(Area Under Curve).Results show that the throughput of the proposed contention window adaptive preliminary adjustment mechanism has been improved by 21.266%compared with that of the LAA/Wi Fi hybrid network without this mechanism.

关 键 词：授权频谱辅助接入技术 WI-FI 混合自动重传请求 机器学习 竞争窗口 吞吐量 

分 类 号：TN801[电子电信—信息与通信工程]